J Nutr Sci Vitaminol, 57, 192–196, 2011

Note Adenosine Triphosphate (AThTP) Inhibits Poly(ADP-Ribose) Polymerase-1 (PARP-1) Activity

Takao TANAKA1, Daisuke YAMAMOTO2, Takaji SATO3, Sunao TANAKA4, Kazuya USUI5, Miki MANABE5, Yui AOKI3, Yasuki IWASHIMA3, Yoshihiro SAITO3, Yoshiki MINO3 and Hirofumi DEGUCHI1

1Organization of Medical Education, and 2Biomedical Computation Center, Osaka Medical College, 2–7 Daigakuchou, Takatsuki, Osaka 569–8686, Japan 3Laboratory of Analytical Chemistry, and 5Laboratory of Pharmacotherapy, Osaka University of Pharmaceutical Sciences, 4–20–1 Nasahara, Takatsuki, Osaka 569–1094, Japan 4Human Health Sciences, Graduate School of Medicine Kyoto University, Syogoin Kawaramachi 53, Sakyou-ku, Kyoto 606–8501, Japan (Received November 5, 2010)

Summary Overactivation of poly(ADP-ribose) polymerase-1 (PARP-1) has been demon- strated to result in various stress-related diseases, including diabetes mellitus. Deficiency of cellular nicotinamide adenine dinucleotide (NAD) content, consumed by PARP-1 to add ADP-ribose moieties onto target proteins, contributes to pathophysiological conditions. Adenosine thiamine triphosphate (AThTP) exists in small amounts in ; however, the function(s) of this metabolite remains unresolved. The structure of AThTP resembles NAD. Recent experimental studies demonstrate beneficial impacts of high-dose thiamine treatment of diabetic complications. These findings have led us to hypothesize that AThTP may modulate the activity of PARP-1. We have chemically synthesized AThTP and evalu- ated the effect of AThTP on recombinant PARP-1 enzyme activity. AThTP inhibited the PARP-1 activity at 10 M, and a structural model of the PARP-1–AThTP complex high- lighted the AThTP binding site. The results provide new insights into the pharmacological importance of AThTP as an inhibitor of PARP-1. Key Words adenosine thiamine triphosphate, poly(ADP-ribose) polymerase-1

Substantial recent experimental studies demonstrate of PARP-1 activation in the pathogenesis of diabetes beneficial impacts of high-dose thiamine on diabetic and its complications has recently been emphasized by complications, such as diabetic retinopathy, diabetic both in vivo and in vitro studies (reviewed in Pacher nephropathy, diabetic neuropathy and diabetic cardi- and Szabó (7), and Szabó (8)). omyopathy (1–4). However, the pharmacological rele- NAD, used as substrate for PARP-1, consists of two vance of high-dose thiamine treatments remains nucleotides joined through their phosphate groups, unknown. with one nucleotide containing an adenine base and Chronic hyperglycemia results in diabetic complica- the other containing nicotinamide. tions in target organs. The pathogenic effect of high glu- Adenosine thiamine triphosphate (AThTP), a new cose is, at least partially, mediated to a significant extent thiamine derivative, was recently identified in Escheri- through increased production of reactive oxygen species chia coli (9), followed by the identification in small and reactive nitrogen species and subsequent oxidative amounts in mouse brain, heart, skeletal muscle, liver stress (reviewed in Evans et al. (5)). Increased oxidative and kidneys (10). AThTP is composed of two molecules, stress activates the nuclear enzyme, poly(ADP-ribose) an adenine base and thiamine, which are joined polymerase-1 (PARP-1). PARP-1 activation depletes its through phosphate groups. The structure of AThTP substrate, nicotinamide adenine dinucleotide (NAD), appears to closely resemble NAD. and also covalently attaches branched nucleic acid-like Although the biological role of AThTP is unknown, polymers of poly(ADP-ribose) to various acceptor pro- the existence of noncoenzyme functions of thiamine teins (reviewed in Kiss and Szabó (6)). A covalently derivatives has been speculated (11–14). attached ADP-ribose polymer, poly(ADP-ribosylation, In the context of 1) structural resemblance of AThTP affects the function of target proteins. The involvement to NAD, 2) the experimental evidence implicating PARP-1 as a causative factor in the pathogenesis of dia- E-mail: [email protected] betes and diabetic complications in vitro and in vivo Abbreviations: AThTP, adenosine thiamine triphosphate; (reviewed in Szabó (8)), and 3) beneficial effects of high- NAD, nicotinamide adenine dinucleotide; PARP-1, poly dose thiamine on diabetic complications, we hypothe- (ADP-ribose) polymerase-1. sized that AThTP could interact with PARP-1 and mod-

192

AThTP Inhibits PARP-1 Activity 193 ulate PARP-1 activity.

Methods Chemical synthesis and purification of AThTP. AThTP was synthesized according to the method described by Bettendorff et al. (9) The compound was initially puri- ® fied by solid phase extraction on a MEGA Bond Elut C18 (Varian Inc., Harbor City, CA, USA) cartridge. After pas- sage of the water solution of the crude compound, the cartridge was washed with water. The compound was Fig. l. Graphic representation of the colorimetric read- eluted with 10 v/v% methanol. The eluted fraction was out of the PARP-1 inhibition curves for AThTP. Each lyophilized, redissolved in water and then purified by point represents the median value from triplicates. size-exclusion on a Bio-Gel® P2 col- umn (2.060 cm) equilibrated with water. The elution profile was followed by reading the absorbance at http://www.chemcomp.com/). 280 nm to detect the presence of AThTP. The AThTP fraction was lyophilized and redissolved in 50 mM Results and Discussion ammonium acetate buffer (pH 7.0). The pure AThTP To test our hypothesis, we chemically synthesized was obtained using an HPLC system equipped with AThTP and the effect of AThTP on PARP-1 enzyme ® a semi-preparative COSMOSIL C18-MS-II column activity was evaluated using recombinant PARP-1 in a (Nacalai Tesque, Inc., Kyoto, Japan) and eluted under cell-free assay. As expected, PARP-1 activity was mark- isocratic conditions using methanol : 50 mM ammo- edly reduced by 10 mM 3-aminobenzamide (approxi- nium acetate buffer (pH 7.0) (5 : 95) at a flow rate of mately 80% inhibition, data not shown). AThTP 3 mL/min. The purity of the preparations was checked showed a dose-dependent effect on PARP-1 activity, pro- by HPLC, MS analysis and NMR. ducing almost complete inhibition at 10 M (Fig. 1). Evaluation of PARP-1 enzyme activity in a cell-free Adenosine thiamine diphosphate (AThDP), thiamine, assay. The effect of AThTP on PARP-1 enzyme activity thiamine diphosphate, and thiamine triphosphate did was evaluated by a HT Universal Colorimetric PARP not inhibit PARP-1 activity at 20 M (data not shown). Assay Kit with Histone-Coated Strip Wells purchased It is interesting to examine a synergistic inhibition from Trevigen (Gaithersburg, MD, USA), following the effect of these materials with ATP and/or ADP, which manufacturer’s instructions. This assay kit measures are not inhibited alone, in detail. A difference of the the incorporation of biotinylated poly(ADP-ribose) onto PARP-1 inhibitory activity between AThTP and AThDP, histone proteins in a 96-well plate. structurally more close to NAD, is thought to be a dif- Molecular modeling of the PARP-1–AThTP complex. ference of interaction with the phosphate to PARP-1 For modeling of the human PARP-1 molecule com- enzyme and a structural flexibility of the phosphate plexed with AThTP, a model of the PARP-1 catalytic moiety in AThTP. domain was prepared by taking the coordinate set, Although the inhibitory concentration of AThTP 1A26 including an ADP molecule (15), from the Pro- (10 M) is much higher than the concentration in tein Data Bank (PDB http://www.rcsb.org/pdb/). Ade- mouse tissues (10), it may be possible that AThTP, syn- nosine and the triphosphate of AThTP were modeled thesized by enzyme, is increased by a mass effect when from this ADP molecule. The PARP-1 binding structure a tissue concentration of thiamine increases. The of the AThTP thiamin moiety was constructed by refer- hepatic concentration of total thiamine is increased by ring to the ABT-888 binding to PARP2 (PDB code: high-dose thiamine and this phenomenon is robust in 3KJD) (16). The thiazole and pyrimidine rings of the streptozotocine-induced diabetic rats (unpublished thiamin moiety were positioned in a similar location to data). Therefore, AThTP is thought to have pharmaco- the imidazole and pyrrolidine rings of ABT-888. Water logical significance as PARP-1 inhibitor. molecules were randomly distributed in a 10 Å shell In this study, we evaluated the PARP-1 enzyme activ- around the complex. After energy minimization using ity in the commerciable available PARP-1 enzyme activ- the MMFF94x force field (17), 200-picosecond (ps) ity kit, which is used for the screening of PARP-1 inhib- molecular dynamics simulations were performed at itors and for measuring the activity of PARP-1 in cell 300 K using a 0.002 ps time step and the NVT method extracts. Unfortunately the precise inhibitory mecha- (18, 19). Finally, the most stable structure during the nism could not be evaluated by this kit. Accordingly, to last 50 ps simulations was optimized by energy minimi- gain further insight, we constructed an initial struc- zation. The potential energy of molecular system after tural model of the PARP-1–AThTP complex by consid- final optimization was 4.99104 kcal/mol including ering the X-ray structures of the PARP-1–ADP ana- the PARP-1 catalytic domain (from Lys662 to logue complex (15) and the PARP2–PARP inhibitor Ser1012), AThTP and 2810 water molecules. All oper- (ABT-888) complex (16), deposited in the Protein Data ations were performed using the package for molecular Bank (http://www.rcsb.org./pdb/) as 1A26 and 3KJD. structure analyses, MOE (Molecular Operating Environ- The energy-minimized structure after 200 ps of molec- ment, Chemical Computing Group Inc., Québec, CA ular dynamics simulations is shown in Fig. 2A as a ster-

194 TANAKA T et al.

Fig. 2. Predicted interaction between human PARP-1 and AThTP. A simulated complex structure of human PARP-1 with AThTP is shown in (A) without hydrogen atoms. AThTP is presented by a light-blue ball-and-stick model, and PARP-1 res- idues adjacent to AThTP are shown by white stick models. Nitrogen, oxygen, sulfur and phosphorus atoms are colored blue, red, yellow and purple, respectively. Each hydrogen bond between PARP-1 and AThTP is indicated by a thin green line with its distance (Å), and the related atoms in PARP-1 are represented by balls. The Ring-stacking interaction between the AThTP thiazole ring and the Tyr907 side-chain is indicated by green circles and dotted lines. An interaction scheme between human PARP-1 and AThTP is shown in (B) with related residues. AThTP is colored light-blue. Hydro- gen bonds, hydrophobic interactions and ring-stacking interactions are illustrated as purple dotted lines, yellow circles and a green circle, respectively.

AThTP Inhibits PARP-1 Activity 195 eo view, and the predicted interaction between PARP-1 The involvement of PARP-1 overactivation has been and AThTP is summarized in Fig. 2B. The initial struc- demonstrated in numerous stress-related diseases ture of AThTP was not provided, but the structure of (reviewed in Pacher and Szabó (7) and Szabó (8)). this simulation resulted in a U-shaped conformation of Accordingly, high-dose thiamine intervention could be AThTP. In this structure, the adenine moiety of AThTP beneficial in treating not only diabetic complications is positioned by hydrophobic interactions with the but also in the treatment of various stress-related dis- Met890 side-chain. Two hydroxyl groups of the AThTP eases. Using an obese rat model, we recently found that ribose moiety are fixed by hydrogen bonds with the high-dose thiamine prevented the metabolic syndrome Glu988 side-chain, and the triphosphate is stabilized by (24) and mitigated the development of hypertension in hydrogen bonds and electrostatic interactions with a spontaneous hypertensive rats (SHR) (25), even though positive charged pocket (His826, Lys903 and the main- the relevance of AThTP still remains to be verified. chains of Tyr985 and Leu986). These interactions are REFERENCES similar to those observed in the ADP analogue complex (15). Ring-stacking interactions were observed between 1)Babaei-Jadidi R, Karachalias N, Ahmed N, Battah S, the AThTP thiazole ring and the Tyr907 phenyl ring. Thornalley PJ. 2003. Prevention of incipient diabetic Furthermore, this phenyl group contributes in fixing a nephropathy by high-dose thiamine and benfotiamine. Diabetes 52: 2110–2120. phosphate group of AThTP via a hydrogen bond. The 2) Brownlee M. 2001. Biochemistry and molecular cell AThTP methyl-pyrimidine enters into a hydrophobic biology of diabetic complications. Nature 414: 813– hole (His862, Leu877, Ile895 and Tyr896), and the 820. amino group of the pyrimidine moiety appears to inter- 3) Hammes HP, Du X, Edelstein D, Taguchi T, Matsumura act with Gln763 and Asp766 through hydrogen bonds. T, Ju Q, Lin J, Bierhaus A, Nawroth P, Hannak D, Neu- Ring-stacking interactions of the thiazole moiety with maier M, Bergfeld R, Giardino I, Brownlee M. 2003. this enzyme were also similar to the X-ray structure of Benfotiamine blocks three major pathways of hypergly- the PARP2–ABT-888 complex (16) and the PARP-1– cemic damage and prevents experimental diabetic reti- PARP-1 inhibitor complexes (20–22). nopathy. Nat Med 9: 294–299. The ADP and ABT-888 binding sites, described 4)Kohda Y, Shirakawa H, Yamane K, Otsuka K, Kono T, above, were reported as ‘‘acceptor’’ and ‘‘donor’’ bind- Terasaki F, Tanaka T. 2008. Prevention of incipient dia- betic cardiomyopathy by high-dose thiamine. J Toxicol ing sites for ADP-ribose elongation by PARP-1 (15). The Sci 33: 459–472. poly (ADP-ribose) chain (acceptor) and an NAD mole- 5)Evans JL, Goldfine ID, Maddux BA, Grodsky GM. 2003. cule (donor) would bind to PARP-1 via the ADP and Are oxidative stress-activated signaling pathways medi- ABT-888 binding sites, respectively. The NAD nicoti- ators of insulin resistance and beta-cell dysfunction? namide moiety is thought to be stabilized by the ABT- Diabetes 52: 1–8. 888 binding site, especially Tyr907 via ring-stacking 6) Kiss L, Szabó C. 2005. The pathogenesis of diabetic com- interactions (16), and many PARP-1 inhibitors have plications: the role of DNA injury and poly(ADP-ribose) been created mainly for binding to this ‘‘donor’’ binding polymerase activation in peroxynitrite-mediated cyto- site. By comparing these binding modes, it appears that toxicity. Mem Inst Oswaldo Cruz 100: 29–37. AThTP in the structure of the PARP-1 complex could 7)Pacher P, Szabó C. 2005. Role of poly(ADP-ribose) poly- interact with both binding sites at the same time. Fur- merase-1 activation in the pathogenesis of diabetic com- plications: endothelial dysfunction, as a common thermore, the AThTP pyrimidine moiety in the above underlying theme. Antioxid Redox Signal 7: 1568–1580. model could interact with an inner hydrophobic hole 8) Szabó C. 2005. Roles of poly(ADP-ribose) polymerase near the binding site of the phenoxypropyl group of a activation in the pathogenesis of diabetes mellitus and quinazoline derivative (PARP-1 inhibitor) (21). How- its complications. Pharmacol Res 52: 60–71. ever, it is also probable that this pyrimidine ring cannot 9) Bettendorff L, Wirtzfeld B, Makarchikov A, Mazzucchelli enter into the hydrophobic hole but covers this hole in a G, Frédérich M, Gigliobianco T, Gangolf M, De Pauw E, similar manner to other inhibitors (22). Angenot L, Wins P. 2007. Discovery of a natural thia- The results herein have reported that AThTP inhib- mine adenine nucleotide. Nat Chem Biol 3: 211–212. ited PARP-1 activity, providing new insights into the 10)Frédérich M, Delvaux D, Gigliobianco T, Gangolf M, Dive pharmacological relevance of AThTP as a regulator of G, Mazzucchelli G, Elias B, De Pauw E, Angenot L, Wins PARP-1. P, Bettendorff L. 2009. Thiaminylated adenine nucle- otides. FEBS J 276: 3256–3268. AThTP is reported to be detectable in the mouse 11) Bettendorff L. 1994. Thiamine in excitable tissues: brain, skeletal muscle, heart, kidney, and liver (10) and reflections on a non-cofactor role. Metab Brain Dis 9: thiamine diphosphate adenylyl transferase, synthesiz- 183–209. ing adenosine thiamine triphosphate, has been charac- 12) Singleton CK, Martin PR. 2001. Molecular mechanisms terized in E. coli (23). We believe that the beneficial of thiamine utilization. Curr Mol Med 1: 197–207. impacts of high-dose thiamine on diabetic complica- 13) Bâ A. 2008. Metabolic and structural role of thiamine tions could result from not only the coenzymatic func- in nervous tissues. Cell Mol Neurobiol 28: 923–931. tion of thiamine but also, at least partly, from a non- 14) Bettendorff L, Wins P. 2009. Thiamin diphosphate in cofactor role for thiamine derivatives in living cells, i.e., biological chemistry: new aspects of thiamin metabo- the inhibitory function of AThTP against PARP-1 activ- lism, especially triphosphate derivatives acting other ity. than as cofactors. FEBS J 276: 2917–2925. 196 TANAKA T et al.

15)Ruf A, Rolli V, de Murcia G, Schulz GE. 1998. The mech- analysis of complexes. Biochim Biophys Acta 1764: anism of the elongation and branching reaction of 913–919. poly(ADP-ribose) polymerase as derived from crystal 22) Miyashiro J, Woods KW, Park CH, Liu X, Shi Y, Johnson structures and mutagenesis. J Mol Biol 278: 57–65. EF, Bouska JJ, Olson AM. 2009. Synthesis and SAR of 16)Karlberg T, Hammarström M, Schütz P, Svensson L, novel tricyclic quinoxalinone inhibitors of poly(ADP- Schüler H. 2010. Crystal structure of the catalytic ribose) polymerase-1 (PARP-1). Bioorg Med Chem Lett domain of human PARP2 in complex with PARP inhibi- 19: 4050–4054. tor ABT-888. Biochemistry 49: 1056–1058. 23) Makarchikov AF, Brans A, Bettendorff L. 2007. Thia- 17) Halgren TA. 1996. Merck molecular force field. I-V. J mine diphosphate adenylyl transferase from E. coli: Comput Chem 17: 490–641. functional characterization of the enzyme synthesizing 18) Berendsen HJC, Postma JPM, Van Gunsteren WF, DiNola adenosine thiamine triphosphate. BMC Biochem 8: 17. A, Haak JR. 1984. Molecular dynamics with coupling 24)Tanaka T, Kono T, Terasaki F, Yasui K, Soyama A, to an external bath. J Chem Phys 81: 3684–3690. Otsuka K, Fujita S, Yamane K, Manabe M, Usui K, 19) Bond SD, Leimkuhler BJ, Laird BB. 1999. The Nosé- Kohda Y. 2010. Thiamine prevents obesity and obesity- Poincaré method for constant temperature molecular associated metabolic disorders in OLETF rats. J Nutr Sci dynamics. J Comp Phys 151: 114–134. Vitaminol 56: 335–346. 20)Ruf A, de Murcia G, Schulz GE. 1998. Inhibitor and 25)Tanaka T, Sohmiya K, Kono T, Terasaki F, Horie R, NAD binding to poly(ADP-ribose) polymerase as de- Ohkaru Y, Muramatsu M, Takai S, Miyazaki M, Kitaura, rived from crystal structures and homology modeling. Y. 2007. Thiamine attenuates the hypertension and Biochemistry 37: 3893–3900. metabolic abnormalities in CD36-defective SHR: uncou- 21)Matsumoto K, Kondo K, Ota T, Kawashima A, Kitamura pling of oxidation from cellular entry accompa- K, Ishida T. 2006. Binding mode of novel 1-substituted nied with enhanced protein O-GlcNAcylation in CD36 quinazoline derivatives to poly(ADP-ribose) polymerase- deficiency. Mol Cell Biochem 299: 23–35. catalytic domain, revealed by X-ray crystal structure